Functions of Interneurons in Mouse Cerebellum

1140 • The Journal of Neuroscience, January 30, 2008 • 28(5):1140–1152

Behavioral/Systems/Cognitive Functions of Interneurons in Mouse Cerebellum

Neal H. Barmack and Vadim Yakhnitsa Neurological Sciences Institute, Oregon Health & Science University, Beaverton, Oregon 97006

The output signal of Purkinje cells is conveyed by the modulated discharge of simple spikes (SSs) often ascribed to mossy fiber–granule cell–parallel fiber inputs to Purkinje cell dendrites. Although generally accepted, this view lacks experimental support. We can address this view by controlling afferent signals that reach the cerebellum over climbing and mossy fiber pathways. Vestibular primary afferents constitute the largest mossy fiber projection to the uvula-nodulus. The discharge of vestibular primary afferent mossy fibers increases during ipsilateral roll tilt. The discharge of SSs decreases during ipsilateral roll tilt. Climbing fiber discharge [complex spikes (CSs)] increases during ipsilateral roll tilt. These observations suggest that the modulation of SSs during vestibular stimulation cannot be attributed directly to vestibular mossy fiber afferents. Rather we suggest that interneurons driven by vestibular climbing fibers may determine SS modulation. We recorded from cerebellar interneurons (granule, unipolar brush, Golgi, stellate, basket, and Lugaro cells) and Purkinje cells in the uvula-nodulus of anesthetized mice during vestibular stimulation. We identified all neuronal types by juxtacellular labeling with neurobiotin. Granule, unipolar brush, stellate, and basket cells discharge in phase with ipsilateral roll tilt and in phase with CSs. Golgi cells discharge out of phase with ipsilateral roll tilt and out of phase with CSs. The phases of stellate and basket cell discharge suggests that their activity could account for the antiphasic behavior of CSs and SSs. Because Golgi cells discharge in phase with SSs,GolgicellactivitycannotaccountforSSmodulation.ThesagittalarrayofGolgicellaxonterminalssuggeststhattheycontributetothe organization of discrete parasagittal vestibular zones. Key words: mossy fiber; climbing fiber; inferior olive; nodulus; uvula; Golgi cell; stellate cell; granule cell; unipolar brush cell

Introduction onto granule cells (Din˜o et al., 2000; Nunzi et al., 2001). Basket Purkinje cell discharge is influenced by climbing and mossy fiber cell inhibitory axon terminals encapsulate Purkinje cell somata afferents. Climbing fibers originate from the contralateral infe- (Andersen et al., 1964; Eccles et al., 1965; Fox et al., 1967). Stellate rior olive and directly evoke low-frequency multipeaked “com- cell inhibitory axon terminals synapse directly on Purkinje cell plex spikes” (CSs) [0.2–8.0 impulses (imp)/s]. Mossy fibers orig- smooth dendrites (Eccles et al., 1965; Fox et al., 1967). Golgi cell inate from vestibular primary afferents and multiple brainstem inhibitory axon terminals contribute to the classic glomerular nuclei. They synapse, often bilaterally, on granule cells whose synapse on granule cells shared by excitatory mossy fiber termi- axons ascend to the molecular layer where they bifurcate and nals (Cajal, 1911; Eccles et al., 1966; Ha´mori and Szenta´gothai, traverse the dendritic tree of Purkinje cells as parallel fibers. The 1966; Fox et al., 1967). Golgi cells also inhibit UBCs (Dugue´ et al., main output signal of Purkinje cells is the modulated discharge of 2005). simple spikes (SSs) (20–60 imp/s) (Brand et al., 1976; Mugnaini, Several “systems level” requirements must be satisfied to un- 1983; Harvey and Napper, 1991; Pichitpornchai et al., 1994). derstand how cerebellar cortical circuitry modulates Purkinje cell Whereas the modulation of CSs can be directly attributed to the discharge. First, the topography of cerebellar afferents must be action of climbing fibers, the modulation of SSs is less well un- characterized so that effective stimuli can be used to drive cere- derstood, although often attributed to mossy fiber discharge bellar activity. Second, parametric control of the stimuli used to (Granit and Phillips, 1956; Thach, 1970; Ebner and Bloedel, 1981; evoke mossy and climbing fiber activity must be established. Armstrong and Edgley, 1988; Nagao, 1989; Lisberger et al., 1994). Without such control, it is difficult to make inferences about Modulatory influences on SSs are not restricted to mossy fiber functional circuitry. Third, independent control of mossy or afferents. Several interneurons synapse on granule cells and/or climbing fiber pathways should be demonstrated. Such control Purkinje cells and may contribute to the modulation of SSs. would enable clear differentiation of their independent contribu- Unipolar brush cells (UBCs) amplify vestibular primary afferent tions. Fourth, anatomical identification of recorded interneurons mossy fiber projections through synaptic feedforward excitation is necessary to infer cell function. Investigation of the circuitry of the vestibulo-cerebellum satisfies these four requirements. Vestibular primary afferent mossy fibers originate from each Received Aug. 28, 2007; revised Dec. 14, 2007; accepted Dec. 17, 2007. of the five ipsilateral vestibular endorgans and project to the ip- This work was supported by National Institute for Deafness and Communicative Disorders Grant DC006668 and silateral nodulus and ventral uvula (Carpenter et al., 1972; Alley National Eye Institute Grant EY018561. We thank Ms. Mary Westcott-Hodson for her expert histological assistance. et al., 1975; Korte, 1979; Kevetter and Perachio, 1986; Gerrits et Correspondence should be addressed to Dr. Neal H. Barmack, Neurological Sciences Institute, Oregon Health & Science University, 505 Northwest 185th Avenue, Beaverton, OR 97006. E-mail: [email protected]. al., 1989; Barmack et al., 1993a; Purcell and Perachio, 2001; Mak- DOI:10.1523/JNEUROSCI.3942-07.2008 lad and Fritzsch, 2003) (see Fig. 1A). Vestibular primary afferents Copyright © 2008 Society for Neuroscience 0270-6474/08/281140-13$15.00/0 also project to the parasolitary nucleus (Barmack and Yakhnitsa, Barmack and Yakhnitsa • Cerebellar Interneurons J. Neurosci., January 30, 2008 • 28(5):1140–1152 • 1141

ipsilateral (left) vestibular primary afferent mossy fibers. Roll tilt onto the left side also increases the discharge of climbing fibers that originate from the contralateral (right) inferior olive and synapse on Purkinje cells in the ipsilateral (left) uvula-nodulus. The ipsilateral (left) primary afferent mossy fiber signal reflects increased depolarization of hair cells in the left labyrinth. The increased climbing fiber discharge reflects decreased excitability of the contralateral (right) primary afferents and the consequent decreased GABAergic input mediated by the right parasolitary nucleus onto the right DMCC and ␤-nucleus. Thus, the primary vestibular afferent signal and the climbing fiber signal originate independently from opposite labyrinths. Despite their separate origins, both vestibular primary afferent mossy fibers and vestibular climbing fibers discharge in phase with ipsilateral roll tilt. Paradoxically, SSs discharge in phase with contralateral roll tilt (Barmack and Shojaku, 1995; Fushiki and Barmack, 1997; Barmack and Yakhnitsa, 2003; Yakhnitsa and Barmack, 2006). Interneurons likely play a role in the phase reversal of SSs with respect to vestibular primary afferent mossy fiber signal. Here we examine the cellular basis of this paradox. We investigate how sinusoidal roll tilt modulates the activity of cerebellar interneurons identified by juxtacellular labeling (Pinault, 1996; Simpson et al., 2005; Holtzman et al., 2006). We measure the vestibularly evoked depth and phase of modulation of Purkinje cell and interneuronal discharge with the aim of discovering how interneurons contribute to the modulation of SSs.

Materials and Methods Anesthesia and surgery. Seventy-four C57BL/6J mice (The Jackson Labo- ratory, Bar Harbor, ME) (weight, 16.0–29.0 g) were anesthetized with intraperitoneal injections of ketamine (60–70 mg/kg) and xylazine (3 mg/kg). Anesthetic depth was evaluated using paw withdrawal and cor- neal reflexes. Body temperature was maintained at 37°C by a servo- controlled heating pad. Mice received supplemental doses of ketamine alone every 20–25 min. Dental acrylic, anchored to the calvarium by four small stainless-steel Figure 1. Vestibular-cerebellar circuitry. Solid lines indicate excitatory pathways. Dashed screws (0–80 ϫ 1/8), affixed an inverted stainless steel screw (4–40) lines indicate inhibitory pathways. A, Roll tilt onto the left side increases vestibular primary between the lambda and bregma sutures. The inverted screw mated with afferent discharge. B, Vestibular primary afferents project to ipsilateral parasolitary nucleus a metal bar that maintained the head rigidly in the center of a three-axis ␤ (Psol), Y-group (Y), and granule cell layer of nodulus. C, Psol projects to ipsilateral -nucleus vestibular rate table. ␤ and DMCC. D, Climbing fibers from the left -nucleus and DMCC project to contralateral (right) The uvula-nodulus was approached by reflecting the muscles overly- ␤ nodulus. E, Y-group projects to contralateral dorsal cap (DC), -nucleus, and DMCC. Amb, ing the occipital bone. An opening (3 ϫ 2 mm) was made in occipital Nucleus ambiguus; cf, climbing fiber; DVN, descending vestibular nucleus; LVN, lateral vestib- bone over folia 7 and 8. A slit (1 mm) was made in the dura mater. The ular nucleus; MVN medial vestibular nucleus; SVN, superior vestibular nucleus; Fl, flocculus; Gc, opening was covered with warm mineral oil. Anesthesia induction and granule cell; icp, inferior cerebellar peduncle; IntP, interpositus nucleus; LCN, lateral cerebellar surgical procedures together lasted 40–60 min. All animal procedures nucleus; MCN, medial cerebellar nucleus; LRN, lateral reticular nucleus; Nsol, nucleus solitarius; were approved in advance by the Institutional animal care and use com- PFl, paraflocculus; Pc, Purkinje cell; pf, parallel fiber; Psol, parasolitary nucleus; SpV, spinal mittee of Oregon Health & Science University. trigeminal nucleus; VI, abducens nucleus; X, dorsal motor nucleus of the vagus; XII, hypoglossal Microelectrode recording. Microelectrodes were drawn from 1.5 mm nerve; 8n, vestibular nerve. capillaries on a two stage puller. They were filled with 2–3% neurobiotin in 0.5 M NaCl and had 10–20 M⍀ tip impedances. Microelectrodes were advanced through folium 8 toward folia 9 and 10 with a digitally con- 2000) (Fig. 1B). The parasolitary nucleus, in turn, conveys a trolled hydraulic microdrive. The maximum penetration distance from purely GABAergic signal to the ipsilateral ␤-nucleus and dorso- the dorsal surface of folium 8 to the ventral surface of folium 10 was 2.2 medial cell column (DMCC) of the ipsilateral inferior olive (see mm. Fig. 1C). Vestibular secondary afferent mossy fibers also project Histological verification of recording sites by juxtacellular labeling with bilaterally to the nodulus and ventral uvula from the descending, neurobiotin. Neurons were labeled juxtacellularly by ejecting neurobiotin medial and superior vestibular nuclei (Brodal and Torvik, 1957; from the micropipette using a stimulus train of positive current pulses Kotchabhakdi and Walberg, 1978; Yamamoto, 1979; Brodal and (1–5 nA, 300 ms, 50% duty cycle) for 1–3 min (Pinault, 1996; Simpson et Brodal, 1985; Thunnissen et al., 1989; Epema et al., 1990). Climb- al., 2005; Holtzman et al., 2006). A “loose patch” was achieved by increasing fibers from the ␤-nucleus and DMCC project to the con- ing current amplitude until the discharge frequency of the recorded neuron was modulated. After juxtacellular stimulation was completed, the tralateral uvula-nodulus (Barmack et al., 1993b; Barmack and micropipette was usually withdrawn to prevent damage to the recorded Shojaku, 1995; Barmack and Yakhnitsa, 2003) (Fig. 1D). neuron. Approximately 80% of the cells stimulated juxtacellularly for During roll tilt about the longitudinal axis, both vestibular Ͼ1.5 min were later identified histologically. Experiments lasted ϳ2.5 h. primary afferent mossy fibers and vestibular climbing fibers are A typical collection of labeled neurons included 6–10 Purkinje cells and modulated. Roll tilt onto the left side increases the discharge of 2–4 interneurons. 1142 • J. Neurosci., January 30, 2008 • 28(5):1140–1152 Barmack and Yakhnitsa • Cerebellar Interneurons

Figure 2. Six interneurons identified by juxtacellular stimulation. Interneurons filled with neurobiotin by juxtacellular stimulation are shown. A, Granule cell. B, Unipolar brush cell. C, Golgi cell. D, Lugaro cell. The arrow indicates four granule cells that were colabeled after juxtacellular staining of the Lugaro cell. E, Stellate cell. F, Basket cell. Note that the scale bar differs for each cell type. gl, Granule cell layer; ml, molecular layer; Pl, Purkinje cell layer.

Approximately 45 min after labeling the last cell in an experiment, the bellar cell type using a bin width of 5 ms and a sampling interval of 500 mouse was deeply anesthetized and perfused transcardially with 20 ml of ms. For Purkinje cell SSs, we used the 5 ms bin width and 500 ms sam- 0.9% NaCl, followed by 100 ml of 4% paraformaldehyde and 0.1% picric pling interval. For CSs we used a 50 ms bin width anda5ssampling acid in 0.1 M PBS, pH 7.4. The brain was removed and fixed overnight at interval. The purpose of comparing median and mean ISIs was to ascer- 4°C. Some brains were serially sectioned at 50 ␮m on a vibratome. Others tain whether a relatively skewed responses were predictive of cell type were cryoprotected, frozen and cut serially on a cryostat at 45 ␮m. Free- (Vos et al., 1999; Simpson et al., 2005; Holtzman et al., 2006). floating sections were processed according to a biotin/streptavidin-HRP Neurons were classified as “responsive” if the second harmonic was protocol. Ͻ50% of the fundamental and if they had a signal-to-noise ratio Ͼ1.5. Photomicroscopy. Labeled neurons were photographed with a digital The second criterion was sensitive to all harmonics and was particularly camera. Images were obtained at three different levels of focus for each applicable to neurons with saturated or “cutoff” responses (Schor et al., neuron. They were combined with a digital image fusion program (Im- 1984). Data are presented as mean Ϯ SD. Polar plots are constructed for age Pro Plus; Media Cybernetics, Bethesda, MD) to obtain an extended each population of identified neurons. The amplitude of the vector in depth of field. these plots indicates mean M. The vector angle indicates the mean ␸ with Vestibular stimulation. The mouse was mounted on a three-axis ves- respect to head position during ipsilateral roll tilt. tibular rate table in the prone position. The rate table oscillated the mouse sinusoidally about a vertical axis (yaw), longitudinal axis (roll), Results and interaural axis (pitch, Ϯ10°, 0.005–0.800 Hz). All recorded neurons ϳ Juxtacellularly identified cerebellar interneurons were tested with a stimulus frequency of 0.2 Hz. Using the technique of juxtacellular labeling we identified two Data analysis. Action potentials were analyzed using Spike 2 software (Cambridge Electronic Design, Cambridge, UK) and displayed in peris- excitatory interneurons (granule cells, UBCs) and four inhibitory timulus histograms, constructed by storing interspike intervals (ISIs) in interneurons (Lugaro cells, Golgi cells, stellate cells, and basket one of 180 bins. Bin width varied with the reciprocal of stimulus fre- cells) (Fig. 2A–F). Each cell type has a distinctive morphology quency: bin width, 1/( f)(180). and histological location that render its identity certain. Granule Spike frequency for each bin was computed from the reciprocal of the cells were identified by their location, their small cell bodies (Ͻ 6 mean ISI as follows: mean rate ϭ 1/(⌺ ISIs in bin)/(bin spike count). ␮m), and their sparse dendrites that have claw-like endings (Fig. This method is preferable to simply adding spike occurrences to bins 2A). UBCs have chalice-like cell bodies (8–10 ␮m) attached to a while discarding information concerning interspike intervals, particu- single, stubby, branched dendrite. They are located within the larly for low numbers of spike occurrences (Barmack and Yakhnitsa, granule cell layer (Fig. 2B). Golgi cells have large cell bodies 2003). Peristimulus histograms included 10–50 cycles of vestibular stim- (10–20 ␮m) with extensive branching of their varicose axon ter- ulation (0.1–0.2 Hz) and were fitted with a cosine function by a least squares method. The phase of responses re: head position was derived minals within the granule cell layer (Fig. 2C). Most often, Golgi from the fitted cosine function: Y(t) ϭ M cos (2␲ft ϩ ␸) ϩ B, where M is cell dendrites extend into the molecular layer. Lugaro cells, like the depth of modulation, B is the average discharge rate, ␸ is the phase Golgi cells, have large cell bodies (10–20 ␮m). However, unlike relative to head position, and f is the stimulus frequency. Golgi cells, Lugaro cell axons are smooth and straight, ending in Interspike interval histograms (ISIHs) were computed for each cere- the granule cell layer. Like the dendrites of Golgi cells, Lugaro cell Barmack and Yakhnitsa • Cerebellar Interneurons J. Neurosci., January 30, 2008 • 28(5):1140–1152 • 1143

tion. Purkinje cells were identified by the classic waveforms of CSs and SSs and by juxtacellular labeling (Fig. 3A–D). A polar plot of responsive Purkinje cells illustrates that the modulation of CSs (red circles) and SSs (open green circles) was antiphasic (Fig. 3D). The population vector for CSs (black arrow) had an M of 0.25 imp/s and a ␸ of 43°. For SSs, M ϭ 1.3 imp/s and ␸ ϭ 209°. Responsive cells were distributed in folia 8–10. The discharge of CSs in 249/269 Purkinje cells increased during ipsilateral roll tilt. The discharge of 20 CSs increased during contralateral roll tilt. Of the 20 con- tralaterally responding CSs, 14 were localized to folia 8 and 9a. The average discharges of CSs and SSs were 0.2 Ϯ 0.3 imp/s and 20.1 Ϯ 10 imp/s respectively (n ϭ 318). The mean and median ISIs were 2.64 Ϯ 0.86 s and 1.79 s (n ϭ 40) for CSs and 0.06 Ϯ 0.03 s and 0.06 s (n ϭ 40) for SSs. The ISIHs for CSs had a tail of responses with longer ISIs, as indicated by a comparison of the mean and me- Figure 3. Purkinje cell CSs respond in phase with ipsilateral sinusoidal roll tilt. A, Juxtacellularly labeled Purkinje cell M428-7 was located 800 ␮m to the right of the midline in folia 10. The arrow indicates a molecular layer interneuron that was colabeled dian ISIs. afterjuxtacellularstainingofthePurkinjecell.B,CSsandSSsweremodulatedbysinusoidalrolltilt.Peristimulushistogramshows Previously, we showed that optimal thatthedischargeofCSs(blackbars)increasedandSSs(graybars)decreasedduringipsilateralrolltilt.Thehistogramswerefitted planes of vestibular stimulation were rep- withcosinefunctions(CSs,redline;SSs,greenline).C,AnISIHforM428-7showsabroaddistributionofintervalsforbothCSsand resented in 400-␮m-wide sagittal zones in SSs. Note the different time scales for CSs and SSs. The CS is illustrated with five and the SS is illustrated with 25 superimposed the uvula-nodulus. Optimal planes for CSs traces. D, A polar plot of the 269/318 responsive Purkinje cells illustrates antiphasic discharge of CSs (red circles) and SSs (open and SSs consistent with stimulation of the green circles). The population vector for CSs (black arrow) had M ϭ 0.25 imp/s and ␸ ϭ 43°. For the population vector for SSs, ipsilateral posterior and contralateral ante- M ϭ 1.3 imp/s and ␸ ϭ 209°. Responses for M428-7 are represented by the filled black (CS) and yellow (SS) circles. For rior semicircular canals were localized to a abbreviations, see Figure 2. medial zone. Optimal planes consistent with stimulation in the plane of the ipsilat- dendrites usually penetrate the molecular layer (Fig. 2D). Stellate eral anterior and contralateral posterior semicircular canals were cells have small cell bodies (8–12 ␮m) and multiple sparsely represented in a lateral zone (Yakhnitsa and Barmack, 2006). We branched (“stellate”) dendrites, confined to the molecular layer characterized as many Purkinje cells and interneurons as rapidly (Fig. 2E). Basket cells have laterally projecting axon terminals as possible, by restricting vestibular stimulation to roll tilt about that envelop adjacent Purkinje cell somata and few, relatively the longitudinal axis, rather than define the “null” and “optimal” unbranched apical dendrites that ascend into the molecular layer planes for each neuron. We confirmed that sinusoidal roll tilt (Fig. 2F). about the longitudinal axis modulated responses of CSs and SSs Juxtacellular stimulation sometimes labeled more than one in Purkinje cells with similar M and ␸ whether the cells were neuron. Granule cells were the most often secondarily labeled activated optimally by preferential stimulation about the axis for interneurons (see Figs. 2D,3A,7A). The distance of secondarily either the anterior or posterior semicircular canals. Although labeled granule cells from the juxtacellularly targeted neuron sug- there was a slight reduction in M for CSs and SSs during stimu- gests that granule cell labeling occurred as a consequence of ret- lation about the longitudinal axis, ␸ remained constant. The vec- rograde uptake of neurobiotin by parallel fiber terminals. This tor differences were not significant (F tests, p Ͼ 0.5; n ϭ 20) (Zar, interpretation accounts for secondarily labeled granule cells 1984). found across a divide of white matter in separate granule cell layers from juxtacellularly stimulated neurons (Fig. 2D). Below, we physiologically characterize five of the six interneu- Mossy fiber terminals rons illustrated in Figure 2. We exclude Lugaro cells from our Although we did not include them among the distinctive neuro- analysis because we have conclusively identified only two. nal cell types encountered in the cerebellar cortex, we successfully labeled four mossy fiber terminals within the granule cell layer Purkinje cell responses (Fig. 4A). The anatomical identity of mossy fibers was established The modulated discharge of CSs and SSs during roll tilt, previ- by their classic rosette terminals. The paucity of our sample does ously observed in Purkinje cells in the rabbit and mouse uvula- not reflect their density within the cerebellum. Mossy fibers com- nodulus, was confirmed (Barmack and Yakhnitsa, 2003; Yakh- prise the largest afferent pathway to the cerebellum. There are nitsa and Barmack, 2006). We recorded from 318 Purkinje cells. ϳ5.0 mossy fibers per Purkinje cell and 0.1 climbing fibers per One hundred and seventy-eight of these Purkinje cells were re- Purkinje cell, yielding an estimate of 50 times more mossy than ported previously (Yakhnitsa and Barmack, 2006). The activity of climbing fibers (Ito, 1984). Rather, it reflects the difficulty in 269 of 318 Purkinje cells was modulated by vestibular stimula- juxtacellularly labeling them without also labeling other cells 1144 • J. Neurosci., January 30, 2008 • 28(5):1140–1152 Barmack and Yakhnitsa • Cerebellar Interneurons

7.1 Ϯ 4.4 imp/s. The mean and median ISIs for mossy fiber terminal discharge were 0.14 Ϯ 0.08 s and 0.12 s (n ϭ 4), respectively.

Granule cells We recorded from a total of 38 granule cells in folia 8–10. The identity of granule cells was established by their small somata (4–6␮m) and their sparse claw-like dendrites (Figs. 2A,5A). It was difficult to isolate single granule cells unless we used elec- trodes with tip impedances Ͼ15 M⍀. The discharge of 28 of 38 granule cells was modulated by vestibular stimulation. Of these modulated granule cells 19 of 28 had type I responses. In 12 of 28 vestibularly driven granule cells, the low-frequency modulation was preceded by a phase-locked high-frequency burst (50–200 Hz) (Fig. 5B,C). We define “bursts” as two or more successive action potentials with ISIs Ͼ50 times shorter than the average ISI during vestibular stimulation (a single burst is shown in Fig. 5B). Granule cell bursts did not occur during every cycle of vestibular stimulation. In the cell illustrated in Figure 5B, bursts appeared in 9 of 11 cycles. Lower discharge rates also occurred within each cycle. In 12 granule cells with both bursts and regular discharges, we measured M and ␸ for the combined responses and for each response separately. We fitted cosine functions to peristimulus histograms that included both the bursts and the “regular discharges” (Fig. 5B1, red cosine function and red vector). We examined bursts only (Fig. 5B2, green cosine function). Bursts had M ϭ 0.6 imp/s and ␸ ϭ 44° (Fig. 5B2, green vector). We excluded bursts and analyzed only the modulated lower rate of discharge (Fig. 5B3, blue cosine function). Under this condition, M ϭ 0.7 imp/s and ␸ ϭ 334° (Fig. 5B3, blue vector). In 5 of 28 vestibularly driven granule cells, we observed only high-frequency bursts. In 11 of 28 vestibularly driven granule cells, we observed only low-frequency modulation (for example of a vestibularly modulated granule cell that lacked bursts, see Fig. 10A,C). The mean discharge of the 28 vestibularly responsive granule cells was 3.3 Ϯ 3.8 imp/s. A polar plot for this population had a vector (red arrow) with M ϭ 0.71 and ␸ ϭ 5° (Fig. 5D). Vestib- ular modulation of granule cells was less than that of mossy fibers. The high-frequency bursting discharge of granule cells was re- flected in the ISIH with two distinct peaks; one reflecting short ISIs, the other reflecting longer intervals (Fig. 5C, red arrowhead). The discharge of granule cells was divided into regular (slow) Figure 4. Mossy fiber terminal type I response. A, Juxtacellularly labeled mossy fiber termi- and burst discharge. The mean and median ISIs for regular gran- nal M425-14 was located 700 ␮m to the left of the midline in folium 9c. B, It had a type I ule cell discharge were 0.21 Ϯ 0.18 s and 0.20 s (n ϭ 33). Granule response. Its discharge increased during ipsilateral roll tilt. The peristimulus histogram (black Ϯ bars) was fitted with a cosine function (red line). C, An ISIH shows a broad distribution of cell bursts had mean and median ISIs of 0.05 0.07 s and 0.05 s intervals.Theactionpotentialisillustratedwith25superimposedtraces.Forabbreviations,see (n ϭ 17), respectively. Figure 2. Unipolar brush cells Unipolar brush cells (UBCs) are found in abundance in the gran- within the granular layer with which their physiological re- ule cell layer, particularly in the uvula-nodulus (Mugnaini and sponses could be confused. Floris, 1994; Din˜o et al., 2000). We recorded from 31 UBCs, The activity of two of four identified mossy fibers was modu- identified in the granule cell layer by their larger somata (8–10 lated by vestibular stimulation. The discharge of one, localized ␮m) and characteristic brush-like endings (Figs. 2B,6A). The 700 ␮m to the left folium 9c (uvula), evinced a type I response by discharge of 27/31 UBCs was modulated by sinusoidal roll tilt increasing its discharge during ipsilateral (left) roll tilt (type I) (Fig. 6B). Ipsilateral roll tilt increased the discharge of 19 of 27 of (Fig. 4B). The vestibularly evoked activity of mossy fibers was the vestibularly modulated UBCs (type I). The population re- well modulated (M ϭ 5.1 imp/s for the type I mossy fiber; M ϭ sponse of UBCs had a phase lead (␸ ϭ 32°) with respect to ipsi- 10.1 imp/s for the type II mossy fiber). The action potential of lateral roll tilt (Fig. 6D). The average discharge frequency of mossy fibers was brief, lasting Ͻ1 ms (Fig. 4C). The average dis- UBCs was 10.0 Ϯ 11.6 imp/s. The average depth of modulation charge rate for the four juxtacellularly identified mossy fibers was (M ϭ 1.2 imp/s) was greater than that of granule cells, but less Barmack and Yakhnitsa • Cerebellar Interneurons J. Neurosci., January 30, 2008 • 28(5):1140–1152 • 1145

mean and median ISIs were 0.23 Ϯ 0.22 s and 0.16 s (n ϭ 23), respectively (Fig. 7C). The extent of dendritic and axonal terminal distribution of Golgi cells is difficult to discern from inspection of photomicrographs (Figs. 2C,7A). Lower power photomicrographs fail to capture the detail of the small diameter, varicose Golgi cell processes. Higher-power photomicrographs fail to capture their spa- tial extent. Therefore we traced the patterns of axonal branching in nine of the better-labeled Golgi cells to gain insight into possible function (Fig. 8). These tracings highlight the most impressive aspect of Golgi cell organization. Their axon terminals have a parasagittal organization. The mean sagittal extent of Golgi cell axons was 650 Ϯ 179 ␮m. The mean mediolateral extent was 180 Ϯ 40 ␮m(n ϭ 9). Because the tissue sections were cut sagittally, we could follow the sagittal trajectories of axons and dendrites in the granule cell and molecular layers. Me- diolateral trajectories could be followed ␮ Figure 5. Granule cells respond to roll tilt with low rates and with bursts. A, Granule cell M399-3 was located 50 m to the right with less precision, because measure- of the midline in folium 9c. B, Vestibularly modulated slow responses were interspersed in 9 of 11 stimulus cycles with high- ments were restricted to increments of frequency bursts. Three peristimulus histograms were plotted and included different combinations of bursts and regular discharge: ␮ burstsandregulardischargewerefittedwithacosinefunction(redline;1);burstsonlywereplotted(cosinefunction,greenline;2); 45–50 m, the thickness of the tissue sec- and regular discharges, without bursts, were plotted (cosine function, blue line; 3). The vectors in B1–B3 indicate the average M tions. Peripheral processes may be in- and ␸ for each condition. C, An ISIH shows two discrete peaks of interspike intervals. The first (red arrowhead) corresponds to the completely labeled juxtacellularly. How- high-frequency bursts. The second consists of more broadly distributed ISIs and corresponds to regular discharge. The action ever, the possibility of more extensive potential is illustrated with 25 superimposed traces. D, A polar plot for the population of 28 vestibularly responsive granule cells branching than we have measured would yieldedavector(redarrow)withMϭ0.7imp/sand␸ϭ5°.ThecombinedresponseofM399-3isindicatedbythefilledblackcircle. not alter the conclusion that Golgi cell axon terminals are sagittally organized. The dendrites of Golgi cells have a than that of mossy fiber afferents. The mean and median ISIs of more varied morphology than do the axon terminals. In most UBCs were 0.12 Ϯ 0.10 s and 0.11 (n ϭ 31), respectively (Fig. 6C). Golgi cells, two to four dendrites emerge as thicker processes from a single locus on the soma. More peripherally, the dendrites Golgi cells narrow and are more varicose in appearance. In fact, peripher- Golgi cells contribute inhibitory axon terminals to cerebellar glo- ally, we could not distinguish Golgi cell dendrites from axons. meruli, the principal sites of mossy fiber termination on granule Consequently, it was not possible to obtain a quantitative esti- cells. Golgi cells also inhibit UBCs (Dugue´ et al., 2005). mate of dendritic organization. For comparison, we measured Extensive axonal branching enables Golgi cells to modulate the asymmetry of the relatively planar dendritic trees of the 6 the activity of thousands of granule cells. Light and electron mi- Purkinje cells. The maximal linear extent of the Purkinje cell croscopic evidence suggests that climbing fibers synapse on Golgi dendritic tree was 190 Ϯ 35 ␮m in the sagittal plane and 120 Ϯ 23 cells (Ha´mori and Szenta´gothai, 1966, 1980; Desclin, 1976; Sugi- ␮m mediolaterally (n ϭ 6). hara et al., 1999). This could allow climbing fiber signals to influ- The uvula-nodulus is organized into at least three physiolog- ence the activity of large numbers of granule cells and UBCs, and, ically defined parasagittal zones (Yakhnitsa and Barmack, 2006). in turn, induce a modulated SS output that is out of phase with Both the CSs and SSs in these zones respond optimally to vestib- mossy fiber–granule cell signaling. ular stimulation about either the axis of the posterior or anterior Golgi cells identified by their relatively large somata (10–20 semicircular canals. The width of these parasagittal vestibular ␮m), extensive axonal branching within the granule cell layer and zones within the mouse uvula-nodulus is ϳ400 ␮m. Golgi cell projection of dendrites toward the molecular layer (Figs. 2C and dendrites and axon terminals could fit within single parasagittal 7A). We recorded from 24 Golgi cells. They had large action zone. Golgi cells could provide one of the mechanisms by which potentials with low average discharge rates (2.7 Ϯ 2.1 imp/s). The similar CS and SS receptive fields are imposed on the discharge of discharge of 18 of these cells was modulated by sinusoidal roll tilt. granule cells, even if those granule cells received overlapping or The discharge of 15/18 of these modulated cells increased during conflicting mossy fiber primary afferent signals. contralateral roll tilt (Fig. 7B). A polar plot for the population of 18 vestibularly responsive Golgi cells yielded a vector (black ar- Stellate cells row) with M ϭ 0.7 imp/s and ␸ ϭ 180° (Fig. 7D). The ISIs of Stellate cells are the only neurons localized exclusively to the Golgi cells, like those of CSs had a distinct low-frequency tail. The molecular layer. We recorded from 47 stellate cells during sinu- 1146 • J. Neurosci., January 30, 2008 • 28(5):1140–1152 Barmack and Yakhnitsa • Cerebellar Interneurons

soidal roll tilt (Figs. 2E,9A). They had dendritic fields of ϳ200 ␮m and somata of 7–10 ␮m in diameter. Stellate cells are abundant. We counted stellate cells in the molecular layer of the mouse nodulus in Nissl-stained, 45-␮m- thick sagittal sections. In each of five sections, we counted stellate cells in a molecular layer that included five Purkinje cells. In five separate samples, we obtained a mean of 22.9 Ϯ 1.5 stellate cells/Purkinje cell. The discharge of 37 of 47 stellate cells was modulated by vestibular stimulation. Of the driven stellate cells, 31 of 37 increased their discharge during ipsilateral roll tilt (Fig. 9B,D). The discharge of the other six stellate cells increased during contralateral roll tilt. Four of these contralater- ally responding stellate cells were localized to folium 9a, a folium with less uniform sagittal organization. A polar plot for the population of 37 vestibularly responsive Figure 6. UBCs respond to roll tilt. A, Juxtacellularly labeled UBC M272-6 was located 350 ␮m to the left of the midline in stellate cells yielded a vector (black arrow) folium 10. B, The activity of M272-6 increased during ipsilateral roll tilt. The peristimulus histogram was fitted with a cosine with M ϭ 2.0 imp/s and ␸ ϭ 11°. function (blue line). C, An ISIH shows a narrow peak of interspike intervals. The action potential is illustrated with 25 superim- The average discharge rate of stellate posed traces. D, A polar plot for the population of 27 vestibularly responsive UBCs yielded a vector (black arrow) with M ϭ 5.4 cells was 10.0 Ϯ 10.9 imp/s. The mean and imp/s and ␸ ϭ 32°. The response of M272-6 is indicated by the filled yellow circle. median ISIs for stellate cells were 0.13 Ϯ 0.10 s and 0.14 s (n ϭ 47), respectively (Fig. 9C). The phase and depth of modulation of the stellate cell population suggest that they could provide an inhibitory input to Pur- kinje cells, in phase with the climbing fibers and vestibular primary afferent mossy fibers, but out of phase with SSs (Fig. 9D). Stellate cells could account for the antiphasic discharge of CSs and SSs.

Basket cells Basket cell somata are found in the lower molecular layer just above Purkinje cells. Their identification depends on the pres- ence of characteristic terminals that envelop Purkinje cell somata and on their relatively long, smooth and sparsely branching dendrites (Figs. 2F,10A). Figure 10 shows a labeled granule cell (M372-3) as well as a labeled basket cell ␮ (M372-4). The discharge of the granule Figure7. Golgicellactivityincreasesduringcontralateralrolltilt.A,JuxtacellularlylabeledGolgicellM338-2waslocated300 mto cell was continuously modulated and had theleftofthemidlineinfolium10.ThearrowindicatesagranulecellthatwascolabeledbyjuxtacellularstainingoftheGolgicell.B,Golgi cellM338-2dischargedinphasewithcontralateralrolltiltvelocity.Itsperistimulushistogramwasfittedwithacosinefunction(redline). no bursts. Its discharge increased during C,AnISIHforM338-2showsabroaddistributionofintervals.Theactionpotentialisillustratedwith10superimposedtraces.D,Apolarplot ipsilateral roll tilt (Fig. 10A,C,D). We were forthepopulationof18vestibularlyresponsiveGolgicellsyieldsavector(blackarrow)withMϭ0.7imp/sand␸ϭ180°.Theresponse unable to maintain juxtacellular stimula- of Golgi cell M338-2 is indicated by the filled green circle. For abbreviations, see Figure 2. tion of the granule cell for more than 1 min. Rather than withdraw the microelectrode, rate of basket cells was 4.8 Ϯ 6.6 imp/s. The mean and median we continued the track. We quickly encountered and labeled the ISIs of basket cells were 0.15 Ϯ 0.06 s and 0.11 s (n ϭ 5), respec- basket cell. Its discharge was modulated by ipsilateral roll tilt (Fig. tively. Basket cells were similar to CSs and Golgi cells in having an 10A,B,D). extended tail of longer ISIs (Fig. 10B). In total, we identified five basket cells, four of whose discharge was modulated by ipsilateral roll tilt. A polar plot for the popula- Topography and polarity of interneuronal responses tion of 4/5 vestibularly responsive basket cells yielded a vector The population responses of two excitatory interneurons (gran- with M ϭ 0.8 imp/s and ␸ ϭ 1° (Fig. 10D). The average discharge ule cells and UBCs) and a major inhibitory interneuron, the stel- Barmack and Yakhnitsa • Cerebellar Interneurons J. Neurosci., January 30, 2008 • 28(5):1140–1152 • 1147

late cell, have similar phases during sinusoidal roll tilt. The population responses of these interneurons were within 20° of ipsilateral roll tilt. Granule cells and UBCs had both type I and type II responses, but with a clear preponderance of type I responses. The responses of Golgi cells differed from those of other interneurons. The population response of Golgi cells was in phase with contralateral roll tilt (␸ ϭ 180°). This response indicates that Golgi cells discharge out of phase with both CSs (␸ ϭ 43°), as well as granule cell-parallel fibers (␸ ϭ 5°), but in phase with SSs (␸ ϭ 209°). Vestibularly evoked activity of granule cells, Golgi cells and stellate cells, but not UBCs was observed as frequently in folia 8 and 9a as it was in folia 9c and 10, although most of vestibular primary and secondary afferent projections are con- centrated in folia 9b,c and 10 (Fig. 11) (Brodal and Hoivik, 1964; Carleton and Carpenter, 1984; Barmack et al., 1993a). These data suggest that vestibular-related activity of granule, Golgi and stellate cells may be more dependent on climbing fiber than mossy fiber–granule cell–parallel fiber activity. Figure 8. Parasagittal organization of Golgi cells. Nine juxtacellularly labeled Golgi cells are drawn in the sagittal plane Discussion through folia 8–10. Each is characterized by profusely branching, fine, varicose axons within the granule cell layer. The distance Antiphasic behavior of CSs and SSs of each cell from the midline is listed. Colors are used to help distinguish the processes of different Golgi cells. For abbreviations, Although the modulation of SSs is almost see Figure 2. universally attributed to activation of mossy fiber pathways (Thach, 1970; Eb- ner and Bloedel, 1981; Armstrong and Edgley, 1988; Nagao, 1989; Kano et al., 1991; Lisberger et al., 1994) several experimental observations are inconsistent with this point of view. First, after the occurrence of a CS, a pause in SSs ensues whose duration exceeds any direct effect of the climbing fiber on Purkinje cell conductances (Granit and Phillips, 1956; Armstrong and Rawson, 1979; McDevitt et al., 1982; Barmack and Shojaku, 1995; Fushiki and Barmack, 1997; Kobayashi et al., 1998; Fukushima et al., 1999; Schiff- mann et al., 1999; Blazquez et al., 2003; Schonewille et al., 2006; Winkelman and Frens, 2006). Second, electrical stimulation of the inferior olive evokes a pause in SSs. This pause persists even when the strength of olivary stimulation is ad- justed so that it does not evoke a CS from the cell from which SSs are recorded (Bloedel and Roberts, 1971). Conse- Figure9. Stellatecellsrespondtoipsilateralrolltilt.A,JuxtacellularlylabeledstellatecellM415-3waslocated250␮mleftof quently, the CS-associated pause in SSs the midline in folium 9c. B, The activity of M415–3 was modulated by ipsilateral roll tilt. The peristimulus histogram was fitted cannot be attributed to a membrane withacosinefunction(redline).C,AnISIHshowsarelativelybroadpeakofinterspikeintervals.Theactionpotentialisillustrated property of the Purkinje cell (Monsivais with 25 superimposed traces. D, A polar plot for the population of 37 vestibularly responsive stellate cells yields a vector (black et al., 2005). More likely, it reflects the arrow) with M ϭ 2.0 imp/s and ␸ ϭ 11°. The response of M415-3 is indicated by the filled green circle. For abbreviations, see action of climbing fibers on one or more Figure 2. classes of cerebellar interneurons. The 1148 • J. Neurosci., January 30, 2008 • 28(5):1140–1152 Barmack and Yakhnitsa • Cerebellar Interneurons

Figure 10. Basket cells respond to ipsilateral roll tilt. A, Photomicrograph shows two separately labeled cells; a basket cell (M372-4) and a granule cell (M372-3) located in folium 10, 100 ␮m to the right of the midline. Discharges of both cells were modulated by ipsilateral roll tilt. B, C, Both cells were fitted with a cosine function (basket cell-red line, granule cell-yellow line). The third trace for each cell is an ISIH. Each shows a relatively broad distribution of interspike intervals. Action potentials are illustrated with 25 superimposed traces. The granule cell lacked the bursts that characterized the discharge of other granule cells. D, A polar plot for the population of four of five vestibularly responsive basket cells yields a vector (red arrow) with M ϭ 0.8 imp/s and ␸ ϭ 1°. TheresponseofbasketcellM372-4isindicatedbythefilledgreencircle.TheresponseofgranulecellM372-3,notincludedinthecomputationofthebasketcellvector,isindicatedbythefilledblue circle. For abbreviations, see Figure 2. phases of vestibularly evoked stellate and basket cell discharge primary afferent mossy fiber signals have been eliminated by a suggest that these inhibitory interneurons could induce unilateral labyrinthectomy (Barmack and Yakhnitsa, 2003). climbing fiber-evoked pauses in SSs. Third, the activity of Sixth, a unilateral microlesion of the ␤-nucleus reduces CS vestibular primary afferent mossy fibers is modulated 180° out and SS modulation in the contralateral uvula-nodulus even of phase with the discharge of SSs and therefore cannot ac- though vestibular primary and secondary mossy fiber affer- count for the modulation of SSs (Barmack and Shojaku, 1995). ents remain intact (Barmack and Yakhnitsa, 2003). These last Fourth, CS and SS modulation in nodular Purkinje cells oc- two points suggest that modulation of primary vestibular curs only during roll-pitch. Modulation does not occur during mossy fiber afferents is neither necessary nor sufficient for the yaw, even though horizontal semicircular canal primary affer- modulation of SSs. ents project onto granule cells in the uvula-nodulus (Kevetter Terminals from single mossy fibers lack topographic spec- and Perachio, 1986; Barmack and Shojaku, 1995; Fushiki and ificity. They are distributed in the granule cell layer both me- Barmack, 1997; Purcell and Perachio, 2001; Newlands et al., diolaterally and rostrocaudally over several millimeters of cer- 2002; Maklad and Fritzsch, 2003). This suggests that vestibular ebellar cortex, in multiple folia (Kevetter and Perachio, 1986; primary afferents not only have the wrong polarity, but that Wu et al., 1999; Purcell and Perachio, 2001; Newlands et al., their synaptic influence is too weak to account for the ob- 2002; Maklad and Fritzsch, 2003). Any remaining topographic served modulation of SSs. Fifth, vestibularly modulated SSs precision conveyed by mossy fiber signals to the granule cell persist in the uvula-nodulus even after all ipsilateral vestibular layer is further reduced mediolaterally by the projections of Barmack and Yakhnitsa • Cerebellar Interneurons J. Neurosci., January 30, 2008 • 28(5):1140–1152 • 1149

Figure 11. Topography and polarity of interneuronal responses to roll tilt about longitudinal axis. The location of each neurobiotin-labeled interneuron is indicated on a two-dimensional map. A, Granule cells. B, Unipolar brush cells. C, Golgi cells. D, Stellate cells. The polarity of interneuronal responses to ipsilateral roll tilt about the longitudinal axis is classified as “in phase” (filled circles) or “out of phase” (open circles). Unresponsive cells are indicated by X. F, Table summarizes responses for all labeled interneurons. G, Sagittal section through the uvula-nodulus shows its correspondence with the two-dimensional representations in A–D.

parallel fibers that course through as many as 500 Purkinje tics of mossy fibers are enhanced by granule cells (Jo¨rntell and cells (Brand et al., 1976; Harvey and Napper, 1991). In con- Ekerot, 2006). trast, the width of a sagittal vestibular zone in the mouse nodulus is ϳ400 ␮m (Yakhnitsa and Barmack, 2006), wide Parasagittal organization of Golgi cells enough to include only ϳ20 Purkinje cells. Contralateral roll tilt increases Golgi cell discharge. Ipsilateral Given this mossy fiber–granule cell circuitry, it is likely that roll tilt decreases it. By discharging in phase with SSs, Golgi every nodular Purkinje cell receives pan-directional vestibular cells impose a low-frequency inhibitory feedback on a large signals mediated by parallel fibers from most, if not all vestibular and heterogeneous population of granule cells and UBCs. The endorgans. In contrast, climbing fibers impose a single dimen- vestibular modulation of Golgi cell activity could reflect a sional input, represented in 400 ␮m wide sagittal zones. The climbing fiber-dependent signal whose sign is reversed by an vestibular specificity represented in the discharge of SSs likely can inhibitory interneuron, say stellate cells. It could also reflect be attributed to cerebellar interneurons driven directly or indi- the same interneuronal activity that modulates SSs. Unlike the rectly by climbing fibers. limited axonal network of UBCs, the distribution of Golgi cell axon terminals is extensive, enabling a single Golgi cell to Excitatory influence of UBCs on granule cells influence the activity of thousands of granule cells arrayed The discharge of UBCs reflects an input mainly from vestibu- within a sagittal zone. lar primary afferent mossy fibers (Rossi et al., 1995; Din˜o et al., The planar organization of Golgi cell axon terminals may ful- 2000, 2001). UBCs amplify vestibular primary afferent signals fill different functions than the planar organization of Purkinje and evoke synaptic feedforward excitation of neighboring cell dendrites. Planar dendrites allow Purkinje cells to be packed granule cells. In a heterogeneous population of mossy fibers, densely while at the same time receiving a maximal array of syn- feedforward excitation by UBCs confers a higher synaptic aptic inputs from parallel fibers. The planar organization of Golgi weighting to those mossy fibers that are driven by vestibular cell axon terminals assures that Golgi cell signals will be restricted stimulation. Possibly the high-frequency bursts of granule to a narrow sagittal array of granule cells (Fig. 8). cells, reflect the influence of UBCs (Hensbroek et al., 2006) The activities of UBCs and Golgi cells may be complementary. (Fig. 5B). Alternatively, these bursts could reflect a rebound UBCs increase the excitatory influence of vestibular primary from Golgi cell inhibition. mossy fiber afferents on granule cells during ipsilateral roll tilt. Although the depth of modulation of single mossy fibers The efficacy of UBCs is locally restricted to the population of faithfully encodes sinusoidal roll tilt, the sensitivity of granule mossy fiber afferents that contact their brush dendrites and the cells to the same stimulus is relatively weak (Figs. 5, 10). Our collection of granule cells within range of UBC axonal terminals. data do not support the view that signal-to-noise characteris- Golgi cells increase inhibition of granule cells during contralat- 1150 • J. Neurosci., January 30, 2008 • 28(5):1140–1152 Barmack and Yakhnitsa • Cerebellar Interneurons eral roll tilt possibly enhancing the dynamic range of granule cell excitatory. Their activity is modulated by vestibular roll tilt and discharge within a sagittal zone. The origin of this Golgi cell dis- out of phase with respect to vestibularly evoked SS modulation. charge is unknown. It may originate from either parallel or The discharge of one class of inhibitory interneuron, Golgi cells, climbing fibers (Ha´mori and Szenta´gothai, 1966, 1980; Desclin, is modulated in phase with SSs and cannot account for decreased 1976; Sugihara et al., 1999). Future experiments in which activa- discharge of SSs after vestibularly evoked CSs. The discharge of tion by either mossy or climbing fibers is selectively reduced two inhibitory interneurons (basket and stellate cells) is modu- should address this question. lated out of phase with SSs. These two inhibitory interneurons Golgi cells have been juxtacellularly labeled previously (Simp- have the correct phase to account for climbing fiber-evoked an- son et al., 2005; Holtzman et al., 2006). In rabbits, the lower mean tiphasic modulation of SSs. A fourth inhibitory interneuron, the discharge frequency of Golgi cells combined with a narrow range Lugaro cell, was not well represented in our recordings. Climbing of ISIs may distinguish Golgi cells from neighboring Purkinje fibers, acting directly or indirectly through basket and stellate cells and obviate the need for juxtacellular labeling. In mice, the cells may impose specific vestibular output signals on SSs within large variability of Golgi cell discharge makes their identification well defined parasagittal zones. This view can be tested directly in using only physiological criteria problematic. future experiments in which the discharge of Purkinje cell SSs is The activity of Golgi cells in cerebellar crus I and II is reduced examined after the unilateral disruption of either climbing fiber by stimulation of both ipsilateral and contralateral trigeminal or or mossy fiber pathways. forelimb afferents (Holtzman et al., 2006). This contrasts with our observation that Golgi cells consistently respond to contralateral roll tilt. Additional experiments will be required to sort References out the effects of Golgi cell discharge on information processing Alley K, Baker R, Simpson JI (1975) Afferents to the vestibulo-cerebellum at the mossy fiber–granule cell synapse. and the origin of the visual climbing fibers in the rabbit. Brain Res 98:582–589. Andersen P, Eccles JC, Voorhoeve PE (1964) Postsynaptic inhibition of cer- Stellate cells contribute to antiphasic modulation of CSs ebellar Purkinje cells. J Neurophysiol 27:1138–1153. and SSs Armstrong DM, Edgley SA (1988) Discharges of interpositus and Purkinje Stellate and basket cells respond in phase with vestibularly evoked cells of the cat cerebellum during locomotion under different conditions. climbing fiber activity. They could inhibit Purkinje cells after a CS J Physiol (Lond) 400:425–445. and thereby account for the antiphasic discharge of CSs and SSs. Armstrong DM, Rawson JA (1979) Activity patterns of cerebellar cortical Stellate cells release GABA onto GABA ␣ receptors on the den- neurones and climbing fibre afferents in the awake cat. J Physiol (Lond) A 1 289:425–448. drites of Purkinje cells, Golgi cells and stellate cells (Pouzat and Barmack NH, Shojaku H (1995) Vestibular and visual signals evoked in the Hestrin, 1997). Our present data do not address the question of uvula-nodulus of the rabbit cerebellum by natural stimulation. J Neuro- whether parallel or climbing fibers are more potent in modulat- physiol 74:2573–2589. ing stellate cell activity. Previously we showed that a unilateral Barmack NH, Yakhnitsa V (2000) Vestibular signals in the parasolitary nu- labyrinthectomy, which eliminates ipsilateral primary afferent cleus. J Neurophysiol 83:3559–3569. mossy fibers, does not decrease ipsilateral SS modulation. Con- Barmack NH, Yakhnitsa V (2003) Cerebellar climbing fibers modulate simple spikes in cerebellar Purkinje cells. J Neurosci 23:7904–7916. versely unilateral inferior olivary microlesions leave vestibular Barmack NH, Baughman RW, Errico P, Shojaku H (1993a) Vestibular pri- primary afferents intact, but reduce modulation of contralateral mary afferent projection to the cerebellum of the rabbit. J Comp Neurol CSs and SSs (Barmack and Yakhnitsa, 2003). These findings sug- 327:521–534. gest that the discharge of climbing fibers is critical for modulation Barmack NH, Fagerson M, Fredette BJ, Mugnaini E, Shojaku H (1993b) of SSs. The present data demonstrate that stellate and basket cells Activity of neurons in the beta nucleus of the inferior olive of the have the correct polarity to account for this inhibitory rabbit evoked by natural vestibular stimulation. Exp Brain Res modulation. 94:203–215. Blazquez PM, Hirata Y, Heiney SA, Green AM, Highstein SM (2003) Cere- Stellate cells receive a signal that is correlated with climbing bellar signatures of vestibulo-ocular reflex motor learning. J Neurosci fiber discharge (Ohtsuki et al., 2004). The synaptic basis of this 23:9742–9751. signal is currently unknown. Like basket cells, stellate cells receive Bloedel JR, Roberts WJ (1971) Action of climbing fibers in cerebellar cortex no somatic synaptic contacts from climbing fibers (Ha´mori and of the cat. J Neurophysiol 34:17–31. Szenta´gothai, 1980). This does not exclude the possibility of axo- Brand S, Dahl AL, Mugnaini E (1976) The length of parallel fibers in the cat dendritic contacts (Scheibel and Scheibel, 1954). Electrical stim- cerebellar cortex. An experimental light and electron microscopic study. Exp Brain Res 26:39–58. ulation of the contralateral inferior olive excites stellate cells with Brockhaus J, Deitmer JW (2002) Long-lasting modulation of synaptic input a synaptic delay of 1–3 ms. This variable delay is independent of to Purkinje neurons by Bergmann glia stimulation in rat brain slices. the amplitude of the extracellularly recorded climbing fiber- J Physiol (Lond) 545:581–593. evoked field potential, implying that the CF-associated response Brodal A, Brodal P (1985) Observations on the secondary vestibulocerebel- might not be attributable directly to a climbing fiber synapse on lar projections in the macaque monkey. Exp Brain Res 58:62–74. stellate cells (Jo¨rntell and Ekerot, 2003). Brodal A, Hoivik B (1964) Site and mode of termination of primary vestibulocerebellar fibres in the cat. Arch Ital Biol 102:1–21. “Spillover” of glutamate from climbing fiber synapses on Pur- Brodal A, Torvik A (1957) The origin of secondary vestibulo-cerebellar fi- kinje cell dendrites could increase the activity of stellate cells bers in cats; an experimental anatomical study. Arch Psychiatr Nervenkr (Szapiro and Barbour, 2007). Alternatively, climbing fibers might Z Gesamte Neurol Psychiatr 195:550–567. influence stellate cells indirectly through a glutamate sensitive Cajal SR (1911) Histologie du syste`me nerveux de l’homme et des vertebre´s. excitation of Bergmann glial astrocytes (Brockhaus and Deitmer, Paris: Maloine. 2002) and astrocytic progenitor NG 2ϩ cells (Lin et al., 2005). Carleton SC, Carpenter MB (1984) Distribution of primary vestibular fibers in the brainstem and cerebellum of the monkey. Brain Res 294:281–298. Overview Carpenter MB, Stein BM, Peter P (1972) Primary vestibulocerebellar fibers Six cerebellar interneurons potentially contribute to the modula- in the monkey: distribution of fibers arising from distinctive cell groups of tion of SSs. Two of these interneurons, granule cells and UBCs are the vestibular ganglia. Am J Anat 135:221–250. Barmack and Yakhnitsa • Cerebellar Interneurons J. Neurosci., January 30, 2008 • 28(5):1140–1152 • 1151

Desclin JC (1976) Early terminal degeneration of cerebellar climbing fibers Kotchabhakdi N, Walberg F (1978) Cerebellar afferent projections from the after destruction of the inferior olive in the rat. Synaptic relationships in vestibular nuclei in the cat: an experimental study with the method of the molecular layer. Anat Embryol (Berl) 149:87–112. retrograde axonal transport of horseradish peroxidase. Exp Brain Res Dinõ MR, Schuerger RJ, Liu YB, Slater NT, Mugnaini E (2000) Unipolar 31:591–604. brush cell: a potential feedforward excitatory interneuron of the cerebel- Lin SC, Huck JH, Roberts JD, Macklin WB, Somogyi P, Bergles DE (2005) lum. Neuroscience 98:625–636. Climbing fiber innervation of NG2-expressing glia in the mammalian Dinõ MR, Perachio AA, Mugnaini E (2001) Cerebellar unipolar brush cells cerebellum. Neuron 46:773–785. are targets of primary vestibular afferents: an experimental study in the Lisberger SG, Pavelko TA, Bronte-Stewart HM, Stone LS (1994) Neural ba- gerbil. Exp Brain Res 140:162–170. sis for motor learning in the vestibuloocular reflex of primates. II. Dugue´ GP, Dumoulin A, Triller A, Dieudonne´ S (2005) Target-dependent Changes in the responses of horizontal gaze velocity Purkinje cells in the use of coreleased inhibitory transmitters at central synapses. J Neurosci cerebellar flocculus and ventral paraflocculus. J Neurophysiol 25:6490–6498. 72:954–973. Ebner TJ, Bloedel JR (1981) Role of climbing fiber afferent input in deter- Maklad A, Fritzsch B (2003) Partial segregation of posterior crista and sac- mining responsiveness of Purkinje cells to mossy fiber inputs. J Neuro- cular fibers to the nodulus and uvula of the cerebellum in mice, and its physiol 45:962–971. development. Dev Brain Res 140:223–236. Eccles JC, Llina´s R, Sasaki K (1965) The inhibitory interneurones within the McDevitt CJ, Ebner TJ, Bloedel JR (1982) The changes in Purkinje cell sim- cerebellar cortex. Exp Brain Res 1:1–16. ple spike activity following spontaneous climbing fiber inputs. Brain Res Eccles JC, Llina´s R, Sasaki K (1966) The mossy fibre-granule cell relay of the 237:484–491. cerebellum and its inhibitory control by Golgi cells. Exp Brain Res Monsivais P, Clark BA, Roth A, Haüsser M (2005) Determinants of action 1:82–101. potential propagation in cerebellar Purkinje cell axons. J Neurosci Epema AH, Gerrits NM, Voogd J (1990) Secondary vestibulocerebellar pro- 25:464–472. jections to the flocculus and uvulo-nodular lobule of the rabbit: a study Mugnaini E (1983) The length of cerebellar parallel fibers in chicken and using HRP and double fluorescent tracer techniques. Exp Brain Res Rhesus monkey. J Comp Neurol 220:7–15. 80:72–82. Mugnaini E, Floris A (1994) The unipolar brush cell: a neglected neuron of Fox CA, Hillman DE, Siegesmund KA, Dutta CR (1967) The primate cere- the mammalian cerebellar cortex. J Comp Neurol 339:174–180. bellar cortex: a Golgi and electron microscopic study. In: Progress in Brain Nagao S (1989) Role of cerebellar flocculus in adaptive interaction between Res, Vol. XXV, The cerebellum (Fox CA, Snider RS, eds), pp 174–225. optokinetic eye movement response and vestibulo-ocular reflex in pig- New York: Elsevier. mented rabbits. Exp Brain Res 77:541–551. Fukushima K, Fukushima J, Kaneko CR, Fuchs AF (1999) Vertical Purkinje Newlands SD, Purcell IM, Kevetter GA, Perachio AA (2002) Central cells of the monkey floccular lobe: simple-spike activity during pursuit projections of the utricular nerve in the gerbil. J Comp Neurol and passive whole body rotation. J Neurophysiol 82:787–803. 452:11–23. Fushiki H, Barmack NH (1997) Topography and reciprocal activity of cer- Nunzi MG, Birnstiel S, Bhattacharyya BJ, Slater NT, Mugnaini E (2001) ebellar Purkinje cells in the uvula-nodulus modulated by vestibular stim- Unipolar brush cells form a glutamatergic projection system within the ulation. J Neurophysiol 78:3083–3094. mouse cerebellar cortex. J Comp Neurol 434:329–341. Gerrits NM, Epema AH, Van Linge A, Dalm E (1989) The primary vestibu- Ohtsuki G, Kawaguchi SY, Mishina M, Hirano T (2004) Enhanced inhibi- locerebellar projection in the rabbit: absence of primary afferents in the tory synaptic transmission in the cerebellar molecular layer of the flocculus. Neurosci Lett 105:27–33. GluRdelta2 knock-out mouse. J Neurosci 24:10900–10907. Granit R, Phillips CG (1956) Excitatory and inhibitory processes acting Pichitpornchai C, Rawson JA, Rees S (1994) Morphology of parallel fibers in upon individual Purkinje cells of the cerebellum in cats. J Physiol (Lond) the cerebellar cortex of the rat: an experimental light and electron- 133:520–547. microscopic study with biocytin. J Comp Neurol 342:206–220. Ha´mori J, Szenta´gothai J (1966) Participation of Golgi neuron processes in Pinault D (1996) A novel single-cell staining procedure performed in vivo the cerebellar glomeruli: an electron microscope study. Exp Brain Res 2:35–48. under electrophysiological control: morpho-functional features of juxta- Ha´mori J, Szenta´gothai J (1980) Lack of evidence of synaptic contacts by cellularly labeled thalamic cells and other central neurons with biocytin or climbing fibre collaterals to basket and stellate cells in developing rat Neurobiotin. J Neurosci Methods 65:113–136. cerebellar cortex. Brain Res 186:454–457. Pouzat C, Hestrin S (1997) Developmental regulation of basket/stellate 3 Harvey RJ, Napper RM (1991) Quantitative studies on the mammalian cer- cell Purkinje cell synapses in the cerebellum. J Neurosci ebellum. Prog Neurobiol 36:437–463. 17:9104–9112. Hensbroek R, Van Beugen BJ, Ruigrok TJ, Simpson JI (2006) Spike modu- Purcell IM, Perachio AA (2001) Peripheral patterns of terminal innervation lation of unipolar brush cells and granule cells in the cerebellum of the of vestibular primary afferent neurons projecting to the vestibulocerebel- awake rabbit. Soc Neurosci Abstr 32:740.2. lum in the gerbil. J Comp Neurol 433:48–61. Holtzman T, Rajapaksa T, Mostofi A, Edgley SA (2006) Different responses Rossi DJ, Alford S, Mugnaini E, Slater NT (1995) Properties of transmission of rat cerebellar Purkinje cells and Golgi cells evoked by widespread con- at a giant glutamatergic synapse in cerebellum: the mossy fiber-unipolar vergent sensory inputs. J Physiol (Lond) 574:491–507. brush cell synapse. J Neurophysiol 74:24–42. Ito M (1984) The cerebellum and neural control. New York: Raven. Scheibel ME, Scheibel AB (1954) Observations on the intracortical relations Jo¨rntell H, Ekerot CF (2003) Receptive field plasticity profoundly alters the of the climbing fibers of the cerebellum; a Golgi study. J Comp Neurol cutaneous parallel fiber synaptic input to cerebellar interneurons in vivo. 101:733–763. J Neurosci 23:9620–9631. Schiffmann SN, Cheron G, Lohof A, D’Alcantara P, Meyer M, Parmentier M, Jo¨rntell H, Ekerot CF (2006) Properties of somatosensory synaptic integra- Schurmans S (1999) Impaired motor coordination and Purkinje cell ex- tion in cerebellar granule cells in vivo. J Neurosci 26:11786–11797. citability in mice lacking calretinin. Proc Natl Acad Sci USA Kano M, Kano M-S, Maekawa K (1991) Simple spike modulation of Pur- 96:5257–5262. kinje cells in the cerebellar nodulus of the pigmented rabbit to optokinetic Schonewille M, Khosrovani S, Winkelman BH, Hoebeek FE, De Jeu MT, stimulation. Neurosci Lett 128:101–104. Larsen IM, Van der BJ, Schmolesky MT, Frens MA, De Zeeuw CI (2006) Kevetter GA, Perachio A (1986) Distribution of vestibular afferents that in- Purkinje cells in awake behaving animals operate at the upstate mem- nervate the sacculus and posterior canal in the gerbil. J Comp Neurol brane potential. Nat Neurosci 9:459–461. 254:410–424. Schor RH, Miller AD, Tomko DL (1984) Responses to head tilt in cat central Kobayashi Y, Kawano K, Takemura A, Inoue Y, Kitama T, Gomi H, Kawato M vestibular neurons. I. Direction of maximum sensitivity. J Neurophysiol (1998) Temporal firing patterns of Purkinje cells in the cerebellar ventral 51:136–146. paraflocculus during ocular following responses in monkeys II. Complex Simpson JI, Hulscher HC, Sabel-Goedknegt E, Ruigrok TJ (2005) Between spikes. J Neurophysiol 80:832–848. in and out: linking morphology and physiology of cerebellar cortical in- Korte GE (1979) The brainstem projection of the vestibular nerve in the cat. terneurons. Prog Brain Res 148:329–340. J Comp Neurol 184:279–292. Sugihara I, Wu H, Shinoda Y (1999) Morphology of single olivocerebellar 1152 • J. Neurosci., January 30, 2008 • 28(5):1140–1152 Barmack and Yakhnitsa • Cerebellar Interneurons

axons labeled with biotinylated dextran amine in the rat. J Comp Neurol Winkelman B, Frens M (2006) Motor coding in floccular climbing fibers. 414:131–148. J Neurophysiol 95:2342–2351. Szapiro G, Barbour B (2007) Multiple climbing fibers signal to molecular Wu HS, Sugihara I, Shinoda Y (1999) Projection patterns of single mossy layer interneurons exclusively via glutamate spillover. Nat Neurosci fibers originating from the lateral reticular nucleus in the rat cerebellar 10:735–742. cortex and nuclei. J Comp Neurol 411:97–118. Thach WT (1970) Discharge of cerebellar neurons related to two main- Yakhnitsa V, Barmack NH (2006) Antiphasic Purkinje cell responses in tained postures and two prompt movements. II. Purkinje cell output and mouse uvula-nodulus are sensitive to static roll tilt and topographically input. J Neurophysiol 33:537–547. organized. Neuroscience 143:615–626. Thunnissen IE, Epema AH, Gerrits NM (1989) Secondary vestibulocerebel- Yamamoto M (1979) Topographical representation in rabbit cerebellar lar mossy fiber projection to the caudal vermis in the rabbit. J Comp flocculus for various afferent inputs from the brainstem investigated by Neurol 290:262–277. means of retrograde axonal transport of horseradish peroxidase. Neurosci Vos BP, Volny-Luraghi A, De Schutter E (1999) Cerebellar Golgi cells in the Lett 12:29–34. rat: receptive fields and timing of responses to facial stimulation. Eur Zar JH (1984) Biostatistical analysis. Englewood Cliffs, NJ: Prentice- J Neurosci 11:2621–2634. Hall.